Revisiting active speakers - HiVi D1010

Back in 2010 I spent many hours tweaking the D1080s with very satisfying results. However now with my Ozone variant DACs I have a far more transparent source than I did then and would like some demo kit which does the DAC justice whilst being fairly compact, portable and not too complex to mod. In 2011 a new smaller and even cheaper model arrived - the D1010 which has undergone some very minor modifications and is now in its mark IV incarnation. Like its older and bigger brother, its also a true active - it uses LM1875 clone ICs for the 100mm bass/mid and a tiny IC amp with a clip-on heatsink powered from a separately regulated supply for the 20mm tweeter. A fairly decent foundation for some extensive hot rodding.

On first connecting a pair of these up to the original portable Ozone (that's the one in the tea canister) I was impressed by the LF soundstage bloom that was reproduced but less than overwhelmed by what happened to the HF. Ragged would be a fair approximation, rather fatiguing to listen to over a longer period. So then I asked myself 'What's the lowest hanging fruit here in terms of mods?'.

Opening up the beastie it turned out that mods to the main amp board are easy because that's got the track side facing uppermost, so no wires need to be disconnected to add on components. The volume/tone/XO board though is another matter entirely - its impossible to remove this because the bass/mid lies right behind the bass pot. I investigated removing the bass/mid but the front baffle may well be glued on, there are no visible screws. So in the end I decided to chop off the spindle of the bass pot in order to get the PCB out.

The crossover board is identical to that used on the D1080, just the amp board is different. It still uses the daft 'tone controls' too so losing the bass pot was no sacrifice at all.

The first mod was a very quick one - 3,330uF across the tweeter amp's supply which resulted in a very worthwhile gain in clarity. Egged on by the high return for very small outlay in time and parts, I moved on to filtering the supply to the 1875s with some LC filtering (hand wound toroidal core inductors on no.28 powdered iron cores). I also have made an attempt at star grounding by cutting and strapping some of the drive unit returns to minimize the loop areas. The XO board has its own winding on the trafo meaning the grounding can be optimized nicely - I added in extra RC filtering on this supply (10R/3,300uF on both rails).

On the XO board (after I summoned up the courage to extract it) I've put in TL431 shunts, and fed these from LM317L CCSs on the amp board. The decoupling has been beefed up to several 220uF/16V. I've also installed 11k resistors from the outputs of all the TL084s to the -ve rail - its to bias the output stages into classA. Impedances of the filtering network were scaled up to ease the loading of the opamps - meaning I reduced the cap sizes by a factor of 5, and the resistors increased by the same multiple. After all these are JFET input amps they should be OK with 10s of kohm impedances right? Most of the filter caps are now C0Gs with the exception of the bass HP filter where I kept the original mylars. The input opamp (the one with attenuation then gain) has been converted to a unity gain follower. I then found that I had a little bit too much gain on hand so went back to the amp board and reduce the power amp gains by 6dB. I could go even lower on the bass/mid amp but the tweeter amp's fixed gain so has an input attenuator.

Overall a very satisfying return for the investment - the two weak points left are that there's some 'sheen' on massed strings and the bass is coloured ('tubby') due to insufficient caps on the supply. There's simply not enough space to fit more in But on the plus side, dynamics are to die for for a 4inch speaker.

I'm tempted to swap out the 1875s (which make the internal heatsink uncomfortably hot) for a single TDA8932 which should work wonders for the bass, but right now they're too much fun to listen to. Fortunately I do have another pair....

Update - seems like I've fixed up the problem of the strings 'sheen' - ceramic decouplers to the rescue right on the tweeter amp's supply pins. No wincing now on my reference Saint Seans Piano Concerto No.1 disc Added some pics to show the internals - the amp board is on the left with three TO220 packages visible at the rear, the two 'LM1875's and the regulator for the tweeter amp. This amp's clip-on heatsink is visible on the left The next pic shows the reason for it running too hot - the heatsink is entirely internal and the fixing method to the back panel isn't particularly thermally conductive. The right-most pic shows the two TL084s and associated Rs and Cs for the XO function, together with the controls. The left-most pot is the one which needs its spindle hacking off.

Update2 - I always get distracted when I undertake one mod and while I'm working on it notice a few other mods that I'd like to do. Hence it gets hard to identify what mods do what. So with the second pair of these speakers I started out with just the tweeter amp decoupling (3300uF and 10*10uF 1206 X7Rs) to remove the HF hash, added 4.7uH to the 12V tweeter reg's output pin to stave off instability from a low ESR cap load, and re-routed the tweeter ground which was paralleled with the bass/mid ground. Then had a listen. The chief difference between this and the hot rodded version is in terms of dynamics, primarily mid/HF dynamics. There's a low level haze behind everything which means no 'jump factor' on transients. Its really like DSD on steroids with barely any soundstage, except at the lowest freqs.

So then my thinking was to start with the source - which means the TL084 opamps. I improved the supply to these which previously had 2 * 220uF after the rectifier diodes, then LM78L08/79L08s. Leaving the regs alone for now I upped the capacitance to 2200uF (the largest caps that'll fit on the footprints provided). This supply is on the amp board then goes through 3 wires to the XO board mounted on the inner right side of the cabinet. On the opamps themselves I put 5k6 resistors between their outputs and the -8V negative rail, thereby biassing them into classA upto 1.4mA or so (1.4mA is the quiescent of each opamp). Result - much better dynamics. This mod costs literally single digit pennies but delivers very worthwhile gains. It seems that 'opamp sound' is largely the result of their classB operation creating a miasma of switching noise on the supplies. A simple resistor biassing their output is a no-brainer.

Next up - more improvements to the opamp supplies. Currently each quad-package has 0.1uF ceramic between each rail and GND, which is the fashionable way to decouple. I replaced the ceramics with 2 * 220uF electrolytics for each. but wired from - to + supply (not to GND). I removed the central 100uF caps, measured their ESR (0.3R) and replaced them with 220uFs with 0.15R ESR. The outputs of the L08 regs for some reason are bypassed with 1000uFs, so I introduced 1R prior to those to give some filtering. I had a listen at this stage - improvements from this stuff were slight (but worthwhile) compared to the first set of mods. I then added an RC stage (10R,2200uF) prior to the regs and I'm not sure I noticed any improvement. However the first pair are more transparent, particularly when the music gets more intense. So I figure the next highest fruit probably is the poweramp supplies and grounding. I shall get winding some more toroidal chokes.....

I managed to get around 3.5m of wire onto each of two Sendust 17mm toroids which showed about 2mH inductance and around 0.5RDC. With 2200uF installed prior to the inductors the noise floor has lowered and stability of the image is improved. The grounding was rationalized at the same time. However massed voices is sounding a tad ragged. I rather suspect to fix this up I'll need to go back to the opamp board and lower the supply impedance still further - perhaps with shunts or large numbers of caps....

Update3 - on a modding roll here so I dug out my old D1080 MkIIs which have the exceedingly nice NXP TDA8947 amp chips inside. I'd previously added lots of yellow wires to re-route the grounds, however with the Ozone DAC I found the sound rather undynamic (I'd not noticed this when I did it because all my DACs were S-D then which have sucky dynamics). I ripped up the wires and re-thought how to do the grounding, deciding to copy the topology used in the smaller ones. That's to feed the opamp supply (a dedicated CT winding on the trafo) only to the XO board, then carry the ground to the amp board on the screens of the output cables. However there's an arrangement to scupper this with this speaker - a power sensing circuit which taakes an AC feed from the opamp supply but uses it with reference to the poweramp ground. In order to eliminate this I went for a voltage multiplier circuit from the main trafo winding, adding a diode and cap. Thus the opamp supp;y is now dedicated to the opamps - which I changed last time to NE5532s. I went through quite a few variations of power supplies to get the dynamics as I like them. In the end I have TL431 shunts which get LC filtered with 2 * 3300uF, inductors being 220uH bobbins. The 5532s have 5k6 output stage biassing resistors. The shunts are fed as for the D1010 with LM317s configured as CCSs.

The poweramp being bridged means I only had to wind one of my Sendust chokes - I've added 22uH bobbins in series with the tweeter amps' supply and plenty of 10uF 1206s. Not sure I've totally tamed the tweeter hash but they're very satisfying to listen to. There's more space internally with these than the D1010s so I'll play around with more caps on the main supply to get the bass tubbiness more under control.

Taking a look inside the 'passive enclosure' this morning I noticed the bass/mid's impedance is marked as 4R. Not a good choice for the (bridged) TDA8937 with its 4A current limit! Lower impedance units give power supplies a harder time too - I'm considering winding an autoformer to take its impedance up to 8R.

Attached Thumbnails